结合多特征的随机森林城市植被提取方法研究
Study on Extraction of Urban Vegetation Based on Random Forest Combining Multiple Features
赵莹 1金丽华 1武建秀 1武丽梅1
作者信息
- 1. 黑龙江第一测绘工程院,黑龙江哈尔滨 150025
- 折叠
摘要
针对城市植被提取研究中国产GF6-WFV影像应用较少,单一特征包含信息量无法更好地获取制备分布信息等问题,本文基于GF6-WFV影像,提取光谱、常用植被指数以及红边植被指数,构建多特征组合的随机森林模型,对城市植被提取进行研究.结果表明:1)多特征组合提取植被时,光谱波段结合常用植被指数以及红边植被指数的提取精度最高,总体精度为87.3％,Kappa系数为0.7386,植被提取精度为80.57％.2)红边植被指数相较于常用植被指数的信息提取中,该指数的相对精度最高.表明本文研究为城市植被信息提取提供了一种具有应用价值的方法.
Abstract
There are few applications of GF6-WFV images produced in China in urban vegetation extraction research,and a single fea-ture containing information cannot better obtain distribution information of vegetation. Based on GF6-WFV images,this article extracts spectrum,commonly used vegetation indices,and red edge vegetation indices,constructs a random forest model with multiple feature combinations,and conducts research on urban vegetation extraction. The results indicate that:(1) When extracting vegetation using multiple feature combinations,the spectral band combined with commonly used vegetation indices and red edge vegetation indices has the highest extraction accuracy,with an overall accuracy of 87.3％,a Kappa coefficient of 0.7386,and a vegetation extraction accu-racy of 80.57％. (2) Compared to commonly used vegetation indices,the red edge vegetation index has the highest relative accuracy in information extraction. Therefore,this study provides a method with good application value for extracting urban vegetation informa-tion.
关键词
多特征/随机森林/城市植被/红边植被指数Key words
multiple features/random forest/urban vegetation/red edge vegetation indices引用本文复制引用
出版年
2024
NETL
NSTL
万方数据